This fix is a bit of a hack, I just moved the declarations into the
'user code' portion of main().
dtcodegen needs to be fixed so as not to generate this type of code in
the first place.
The symbol hp-ux is used to tag files for installation on HP machines
in the UDB databases. To ensure the correct release streams are
defined, these symbols (hp-ux, linux, freebsd, etc) are undef'd so
that they are not evaluated in the context of db file generation by
the preprocessor.
A change in the FreeBSD 10 patch disabled this, possibly due to
'hp-ux' being an invalid cpp symbol on FreeBSD 10 machines, which now
use clang by default.
Undefining these cpp symbols is actually required, otherwise linux,
for example, is defined as a '1' in the generated databases, which
will not match the release stream name 'linux', therefore all of the
files tagged as 'linux' are never installed.
To fix this, revert the change made in the FreeBSD 10 patch, and
rename all occurances of 'hp-ux' to 'hpux' in the UDB files to avoid
the potential for cpp trouble when hp-ux is specified.
As a result of this change, 'hpux' is now the name of the release
stream for hp machines, not 'hp-ux'.
This would allow different locales to be specified for
programs/localization (DtLocalesToBuild) and doc/
(DtDocLocalesToBuild).
This would be useful for 'partial' translations like the Greek el_GR
locale, where message catalogs, actions, or other items are localized,
but documentation like the help system and dtinfo are not.
The cause of the bug was that X*DrawImageString draws background
according to the extents of the given string, not to extents of the font set,
which determine terminal line height.
Now, when such a situation is detected, the background is cleared before
drawing the characters.
The patch converts desktop_grid[] from Boolean array into array of
counters of objects, which are placed on cells. When object is
placed on /removed from the screen the counter gets incremented/decremented.
The panel registration code rather stupidly assumed that display
size is always 1280x1024 pixels. Because of this, depending on screen
size, the panel could be registered somewhere in the center of the screen
or completely or partially beyond of it.
The panels were registered only on startup, not those, which were added
from UI.
The fix moves panel registration into separate routine and removes assumption
about display size. The fix yet is not complete since it still makes assumption
about panel's geometry: from dtfile there is still no way to find out
dinamically the size of panel and it's location.
On small screens segfaults could be also triggered without any icons on dtfile
startup if dtwm panel (or part of it) was registered beyond the screen when
RegisterInGrid() was called by InitializeDesktopGrid().
The patch also makes grid registration work for large objects (larger than
2 cells in any direction, like dtwm panel or icon with long file name).
Previously only rectangle vertices were registered.
There was a check in linux.cf for the Linux libc version that didn't
work, since these were never set anywhere. Presumably current Xorg
imake sets these, but since we aren't using that... (yet)
As a result, the build assumed that thread-safe API's were not
supported. Setting the default libc major version to '6' allows the
proper build to take place. libc5 was never threadsafe and no one has
probably used it in over 10 years.
As a result, some earlier porting fixes that referenced '__fds_bits'
on linux systems had to be reverted as these are not valid in an MT
environment. They are also not neccessary when building in such an
environment, as the normal 'fds_bits' works fine.
This patch defintely needs testing on many linux systems.
On some systems with libtcl installed, it might be linked into instant
rather than the locally provided version. This can result in random
coredumps.
Seen on Ubuntu 12.04, x86_64.
Now, we force linking with ../tcl/libtcl.a.
In init() there was code iterating over all of the possible file
descriptors in a svc_fdset. fdsets are limited to FD_SETSIZE. This
caused coredumps on FreeBSD 10, and possibly other hidden issues.
Moving to poll(), rather than select() would be better, but is a bigger
job. For now, just limit to the FD_SETSIZE that select() requires.
With this patch, it is no longer neccessary to create the
/imports/x11/include/X11 symlink...
Also, remove the include of xfree86.cf from linux.cf, and define our
own DefaultCCOptions. Setup to use only -ansi, not -pendantic.
gcc can generate dependencies, so use that instead of building and
depending on the rather delicate and archaic 'makedepend'.
This fix removes makedepend from being built or used on linux, and
instead uses gcc to generate dependency data in .depend files.
I think pretty much every platform that uses gcc should do this. I am
not sure if other compilers (intel, clang) can do this though.
libcompat and header files will be gone in the upcoming 5.5 release, so we
cannot use the ftime(3) interface there. Fall back to the SVR4 version of the
code, but use the tm_gmtoff field of struct tm to get the timezone.
Unfortunately, that is not portable (because I really really would like to
remove that old and crummy struct timeb from the public API).
In several places, variables were XFree()'d, but then used in system()
commands later. This would cause various issues - on my system, I'd
get:
sh: 1: Syntax error: EOF in backquote substitution
system for rm failed; exiting...
while building the guides.
Mostly missing headers, explicit parentheses and new prototypes.
Some Caveats:
* I haven't compile-tested the SVR4 getpty file, it might need another tweak
* There were operator precedence bugs in TermPrimCursor.c and TermPrimRender.c
(^ vs. !=). This might change behaviour, but at least I haven't experienced
any crashes ...
* This adds a little more dependencies for include ordering, but unless we
want to play the "headers that include headers that include headers..." game,
this is unavoidable.
This reverts commit 8a8619bfa8.
More work will need to be done to use tirpc on 64bit systems. It
works 'accidentally' on 32b systems. The issue is that tirpc
includes *must* be used, and there is some work required to properly
support this (like proper include paths, proper definition of XDR,
etc).
So for now, we revert this until that work can be completed and tested,
otherwise 64b linux builds are likely to have problems.
Currently, mp_rpc_server.C tries 538 million ports to acquire an
available transient rpcbind port number. This is bad when rpcbind is
running in secure mode (and you are not using tirpc) - Xsession will
'hang' at the dthello (blue) screen filling up your error logs with
RPC errors.
Now, just try +- 50 (for a total of 100 ports) before bailing. The
dthello 'blue screen of death' is the most common problem in starting
CDE when rpcbind isn't set up properly. This should at least not
cause the appearance of a 'hang'.
This adds a basic library and support to dtsession and dtlogin to
support Xinerama/Twinview, where multimple monitors are used to make
up an X11 screen.
The main goal here is to draw dialogs and such centered on a monitor,
rather than spread out over multiple monitors.
Might need to add sorting - as on my test system, what I would
consider monitor 0, appears to actually be monitor 1. So a sort might
need to be added to sort the screens according to increasing x and y
offsets so it make sense to a user.
Also, this library is built statically and not documented. Maybe it
could be 'filled' out and refactored/redesigned in the futre if need
be and suppoerted.
It is enabled via a define, CDE_USEXINERAMA in site.def. It's a very
simple lib, so I do not expect any issues with the BSD's - it should
build and work fine, assuming your X server has the XINERAMA
extension, which I think pretty much all of them do at this point.
On OpenBSD, the 'S' option to malloc(3) enables guard pages (among other
things). This loop could have triggered this trap when reading beyond the
buffer. Also, the whole "while(*ip)" construct was based on the assumption that
the memory after the string is always zero-filled.
This reverts commit 44e384aedb.
This code is actually needed. If svcfd_create() is not available, it
should be fixed only for those systems that it affects.
This has the effect of not performing a tt call each time in
ResolveLocalPathName() if we're on the local host anyway. Drastically reduces
dtfile startup time.
Use Adobe Helvetica as the sans serif user
interface font (among others, dtlogin, front
panel buttons, menu titles) instead of Lucida.
While there, move X Consortium comment back
to the top.
This change applies only to FreeBSD.
libDtHelp is unable to read SDL help files
with -ftree-store-ccp optimization which
is enabled by -O2 on gcc 4.2.1.
GifUtils.c and decompress.c didn't work
properly with -ftree-store-ccp enabled.
GifUtils.c was repaired by fixing
those warnings:
GifUtils.c: In function 'create_pixmap':
GifUtils.c:1093: warning: return makes integer from pointer without a cast
GifUtils.c:1110: warning: return makes integer from pointer without a cast
GifUtils.c:1215: warning: return makes integer from pointer without a cast
GifUtils.c: In function 'gif_to_pixmap':
GifUtils.c:1242: warning: return makes integer from pointer without a cast
decompress.c didn't generate warnings, but the
only effect of the -ftree-store-cpp was to introduce
this change:
addq $1, %rax
movq %rax, (%rbx)
.L90:
- cmpl $157, %edx
+ cmpl $-99, %edx
jne .L86
movl 8(%rbx), %eax
subl $1, %eax
Which corresponds to this source code:
bufioI.h
57 #define BufFileGet(f) ((f)->left-- ? *(f)->bufp++ : (*(f)->io) (f))
42 int (*io)(/* BufFilePtr f */);
decompress.c
53 #ifdef NO_UCHAR
54 typedef char char_type;
55 #else
56 typedef unsigned char char_type;
57 #endif /* UCHAR */
58
59 static char_type magic_header[] = { "\037\235" }; /* 1F 9D */
131 if ((BufFileGet(f) != (magic_header[0] & 0xFF)) ||
132 (BufFileGet(f) != (magic_header[1] & 0xFF)))
133 {
134 return 0;
135 }
BufFileGet() returns (int), so the (unsigned char) constants
got promoted to (int) with sign extension; therefore constant
157 decimal (0x9D) became -99 decimal, sign extended
(0xffffff9D), and the comparison was always false.
Tested using:
$ gcc -v
Using built-in specs.
Target: amd64-undermydesk-freebsd
Configured with: FreeBSD/amd64 system compiler
Thread model: posix
gcc version 4.2.1 20070831 patched [FreeBSD]
Running on:
FreeBSD 10.0-CURRENT (r240948M)
built Wed Sep 26 23:33:08 CEST 2012
dtcreate crashed on 64-bit system when clicking
"Find Set.." button.
Crash happens in libXm:
new_w=0x805db4300, args=0x7fffffffb430, num_args=0x7fffffffb3dc)
at Form.c:1955
$1 = {att = {{type = 4 '\004', w = 0x805db3700, percent = 0, offset = 0,
value = 0, tempValue = 0}, {type = 1 '\001', w = 0x0, percent = 0,
offset = 10, value = 0, tempValue = 0}, {type = 3 '\003',
w = 0x805db3700, percent = 0, offset = 0, value = 0, tempValue = 0}, {
type = 3 '\003', w = 0x800000000, percent = 0, offset = 10, value = 0,
tempValue = 0}}, next_sibling = 0x0, sorted = 0 '\0',
resizable = 1 '\001', preferred_width = 0, preferred_height = 0}
(...)
at icon_selection_dialog.c:1768
1767 /* Creation of icon_scrolled_win */
1768 icon_scrolled_win = XtVaCreateManagedWidget( "icon_scrolled_win",
1769 xmScrolledWindowWidgetClass,
1770 icon_selection_dialog,
1771 XmNscrollingPolicy, XmAUTOMATIC,
1772 /* XmNnavigationType, XmTAB_GROUP, */
1773 XmNx, 282,
1774 XmNy, 84,
1775 XmNscrollBarDisplayPolicy, XmAS_NEEDED,
1776 XmNrightOffset, 10,
1777 XmNrightAttachment, XmATTACH_FORM,
1778 XmNtopOffset, 0,
1779 XmNtopWidget, icon_container_label,
1780 XmNtopAttachment, XmATTACH_WIDGET,
1781 XmNleftOffset, 0,
1782 XmNleftWidget, icon_container_label,
1783 XmNleftAttachment, XmATTACH_OPPOSITE_WIDGET,
1784 XmNbottomOffset, 10,
1785 XmNbottomWidget, XmATTACH_NONE,
1786 XmNbottomAttachment, XmATTACH_WIDGET,
1787 NULL );
What happens here is that ConstraintInitialize receives
four constraints, the last one is this:
{ type = 3 '\003', /* XmATTACH_WIDGET */
w = 0x800000000, /* malformed XmATTACH_NONE ???
percent = 0,
offset = 10, /* specified as XmNbottomOffset */
value = 0,
tempValue = 0}
XmATTACH_* values are defined in <Xm/Xm.h> as follows:
505 enum{ XmATTACH_NONE, XmATTACH_FORM,
506 XmATTACH_OPPOSITE_FORM, XmATTACH_WIDGET,
507 XmATTACH_OPPOSITE_WIDGET, XmATTACH_POSITION,
508 XmATTACH_SELF
509 } ;
What is not clear to why XmATTACH_NONE - which should be (int)0 -
becomes 0x800000000 - looks like a 64 bit bug somewhere.
Providing a long value on None (0L) as in this change fixes the
problem.
I understand is that it possible to use such an "empty" widget
is to create additional space at the bottom of the newly created
"icon_scrolled_win".
What needs to be clarified - shouldn't be such an (int) value be
automatically promoted to (long) (or XtArgVal, XtPointer, ...)
and preserve the value 0? Lots of parameters seem to be
passed as ints (for example dimensions) and they do not
appear to cause any trouble.
XmPrivate must be generated manually. For this you work, you must
have a freshly compiled openmotif tree, and MLIBSRC must be pointing
to it.
Otherwise, it's possible during the includes phase for an attempt to
be made to regenerate this file, which will fail on the vast majority
of systems out there.
So, to regenerate,
cd include/Xm
rm XmPrivate.h
make XmPrivate.h
- Fix missing prototypes
- Fix some 64-bit related problems (XtVaGetValues)
- Fix crash on dtcreate startup in create_applicationShell1()
- Add XmeFlushIconFileCache() prototype from <Xm/IconFileP.h>
When applying a patch, "git am" strips
trailing whitespace, although they are
present in the git formatted-patch.
This way the committed file will be
slightly different than the file re-generated
by extractprototype.h
It shouldn't hurt, but next run of
extractprototype.h will add trailing spaces
again and the resulting diff on XmPrivate.h
will include more changes than actually
needed.
This may break some viscious circle after
applying the patch, so enabling regeneration
on LinuxArchitecture again.
This patch does not add XmeFlushIconFileCache()
needed by dtcreate.
__rpc_xdr is no longer available on FreeBSD 10.
(XDR is typedef'd as "struct XDR" and not "struct __rpc_xdr").
By the way, why did we ever need this? Probably
it should be removed. Leaving for __OpenBSD__ for now.
This is required by recent changes to the stack protector code in gcc,
generating references to __guard_local instead of __guard, defined in
crtbeginS.o.
This is a temporary solution; strictly speaking, we shouldn't invoke ld
directly at all but use cc instead.
Install black and white and 128-color custom logos for FreeBSD.
The FreeBSD logo based on the artwork provided by The FreeBSD Foundation:
http://www.freebsd.org/logo/logo-basic.png
The mark FreeBSD is a registered trademark of The FreeBSD Foundation
is are used by Common Desktop Enviroment (CDE) with
the permission of The FreeBSD Foundation.
The FreeBSD Logo is a trademark of The FreeBSD Foundation and is used
by Common Desktop Environment (CDE) with the permission of
The FreeBSD Foundation.
Use of logo subject to Trademark Usage Terms and Conditions:
http://www.freebsdfoundation.org/documents/Guidelines.shtml
The following patch gets the Lucida Sans font working on my FreeBSD system. Before applying this, title bars and menu bars are displayed in the "-misc-fixed" font. This is on FreeBSD 9.0-RELEASE-p3.
On Mon, 24 Sep 2012, Jon Trulson wrote:
> On Tue, 25 Sep 2012, Marcin Cieslak wrote:
>
> Applied.
I'm sorry - it turns out this one does not fully
work as expected.
This one is better (it's relative to the old master)
- so it might cause a conflict:
Fixes the following warning:
In file included from ../../../imports/x11/include/X11/Xutil.h:54,
from ../../../imports/x11/include/X11/Intrinsic.h:54,
from Action.c:64:
../../../imports/x11/include/X11/keysym.h:49:1: warning: "XK_MISCELLANY" redefined
<command-line>: warning: this is the location of the previous definition
<keysym.h> which includes all key symbols and loads <keysymdef.h>
is automaticlly included by the X Toolkit.
This patch removes #include <keysymdef.h> whenever not needed,
and adds #define XK_MISCALLANY in the source code where required.
Fix this warning:
RFCTransport.C: In function 'long unsigned int writeToFileDesc(const char*, int,
__va_list_tag*)':
RFCTransport.C:91: warning: 'DtMailBoolean' is promoted to 'int' when passed thr
ough '...'
RFCTransport.C:91: warning: (so you should pass 'int' not 'DtMailBoolean' to 'va
_arg')
RFCTransport.C:91: note: if this code is reached, the program will abort
Having a difficult choice between unplasant
cast to get a void * into an enumeration type
and "Something's wrong here" double cast
I decided for the latter.
At least it does not crash when the legal
value of zero is passed as the argument.
Avoid overwrite of local variables when using
short (int, etc.) types with XtVaGetValues().
Cast XtPointer using (XtArgVal) without
the need to use C99 <stdint.h> and friends.
Fix SIGSEGV because of implicit declaration
of _XmStringUngenerate.
The error message reported to the user was:
TT_ERR_PROCID The process id passed is not valid.
XtArgVal should be a type that encompasses XtPointer
and long integer types. In the X.org implementation
it is currently defined as (long).
Don't use (unsigned int *) instead of (Window *).
Use /usr/local/bin/lpq from CUPS for FreeBSD
for now; this prevents immediate dtprintinfo
crash.
In the future we should handle both built-in
/usr/bin/lpq as well as CUPS /usr/local/bin/lpq
output in separate functions.
Code to support CUPS should probably shared between
other operating systems.
When asking for data using XtVaGetValue()
make sure that there is enough place for
the return value (which is sometimes XtPointer).
Providing pointer to (int) is not enough.
Cast XtPointer into requested int types
directly, which unfortunately introduces
compilation warning:
cast from pointer to integer of different size
Add missing prototypes and header files to the dtfile
code in order to move closer towards 64-bit compatibility.
Extract the following functions from Motif internal headers:
_XmGetWidgetExtData
_XmRecordEvent
_XmStringUngenerate
_XmTextFieldSetDestination
_XmGetActiveTopLevelMenu
Extract manually prototypes of the obsolete Motif interface:
_XmHighlightBorder
_XmUnhighlightBorder
Remove XmPrivate.h if extractprototype.awk fails
Make the following header files available via -I:
codelibs/boolean.h
codelibs/pathutils.h
codelibs/shellutils.h
and remove shellutils.h from dtwm directory.
* Revert changes to programs/dtsearchpath/libCliSrv/UnixEnv.C
introduced by c3cb5b8aa6
that could have produced disappearing Application Manager
icons on FreeBSD
* Introduce SearchPath:useSystemPath() virtual method to tell
dtsearchpath to leave some environment variables alone.
It is currently overriden for FreeBSD only if the MANPATH
is empty (system default). Other operating systems
might want to override it if they prefer to have distribution
specific control of a search path in effect.
* Symlink /usr/dt/share/man to /usr/dt/man for FreeBSD
This allows dtsearchpath to actually include /usr/dt/man
in the MANPATH when MANPATH override is in effect.
The dtexec code assumes that fstat reports pipe's readable chars.
Linux always reports 0 for st_size of a pipe.
Instead read one character when select reports readable.
Note EOF when select says readable but read returns 0.
- build shared libraries with major number only (libtt.so.2)
- don't build dtlogin and dtinfo
FreeBSD support for the installer:
- work around awk issue
- create post_install FreeBSD scripts
- install only cmsd on /etc/inetd.conf
skip dtspc and ttdbserver for now
Note to users: please check
if you are affected by awk bug
in udbParseLib.awk if you can.
Currently on Linux, you must run rpcbind in insecure mode (-i) in
order for ttsession to register with rpcbind (the portmapper). This
is because, on most systems, libc contains an older sun-based rpc
library embedded within it. The sun-rpc code does not know how to
authenticate with rpcbind when run as a non-privileged user.
Using libtirpc, ttsession can register with rpcbind without requiring
it to be run in insecure mode. You must have the libtirpc-dev, or
equivalent package installed to use it.
If you want to try this:
- install libtirpc-dev or equivalent
- if your rpcbind process is already running in insecure mode (-i
option), remove that option and restart it.
- edit config/host.def, (create if it doesn't exist) and add:
#define HasTIRPCLib YES
- rebuild CDE (make World). It's probably a good idea to remove
/usr/dt/* beforehand to avoid contamination.
Some linux systems seem to incorporate tirpc directly into libc (as
all of the BSD's do AFAIK) so this may not be needed.
I know that at least on Ubuntu systems defining HasTIRPCLib to YES is
required in order to run rpcbind without -i.
According to the spec, blank lines in message catalogs or lines
beginning with '$ ' are valid comments.
However, there were many cases where lines in the message catalogs
contained just a single '$', without the required space after it.
Under linux, this caused 126766 error lines (in my builds) of the
form:
... unknown directive `': line ignored
This also causes gencat to exit with a non-0 exit code. Even though
gencat says it ignores the line, it really doesn't.
An early porting change to programs/localized/util/merge.c was made to
ignore this return value on linux. This hack has now been removed.
Build logs are a lot smaller and cleaner now.
this means the compiler treats them as system headers and does not give you
excessive warnings from them. This is used because X11 does not like the
-ansi and -pedantic warnings that CDE is compiled with.
Having this dependancy in here is a problem. Depending on how the src
is packaged, or, unpacked, an attempt could be made to regenerate this
file, which cannot succeed unless imports/motif is setup to point to a
compiled motif tree.
This file can be manually regenerated by removing XmPrivate.h, and then
doing a 'make XmPrivate.h' in the include/Xm dir.
- Add missing includes and prototypes
- Improve type compatibility
- Use <Xm/XmPrivate.h> for unofficial libXm headers
With this patch, dtpad no longer crashes on startup
on FreeBSD/amd64 because of a truncated 64-bit pointer.
- Improve pointer/int compatibility
- Include unpublished Dt headers in Dt
- Use <Xm/XmPrivate.h> for unpublished Motif functions
There are still warnings left generated because
ElementValue.parsed_value should really be a union.
There are also some warnings left because of XtPointer
casting and some unused variables and functions.
We need ANSI C prototypes of certain Motif
functions that are not published in the official
header files.
<Xm/XmPrivate.h> header file contains
the prototypes extracted from the Motif source.
To re-create <Xm/XmPrivate.h>:
1) Make sure you have sources of libXm
(lib/Xm directory of the Motif distribution)
accessible via imports/motif/lib/Xm
2) rm include/Xm/XmPrivate.h
3) make includes
Tested this change with both gawk and nawk and it worked fine. If
the extra escape character is present and gawk is used, you'll get
warnings from gawk telling you it's ignoring the escape sequence
and just treating it as the character to begin with.
This patch removes instances of hardcoded
invocation of /bin/ksh and allows to
replace it with, for, example,
/usr/local/bin/ksh93
Also "ksh93" is accepted whenever "ksh" is.
Tested using the following /bin/ksh:
----8<----
WHAT=`ps -o command= -p $PPID`
msg="Something tried to call /bin/ksh: $PPID: $WHAT"
print -u2 "$msg"
logger user.warn "$msg"
exit 99
----8<----
(Warning: first two lines are FreeBSD specific)
Scripts from Makefiles should now be executed either
with
$(KORNSHELL) korn-shell-script
or
$(SHELL) bourne-shell-script
therefore #!/bin/ksh has not been changed everywhere.
/usr/dt/bin/ scripts have been converted (e.g. Xsession)
Whenever possible Imake and CPP facilities have been used.
For C and C++ programs KORNSHELL needs to be defined to
"/path/to/your/ksh" (with quotes) so that it can make
a valid C constant.
Therefore, when adding KORNSHELL to Imakefile for C files,
you have to add
CXXEXTRA_DEFINES = -DKORNSHELL=\"$(KORNSHELL)\"
or similar (for example, see programs/dtprintinfo)
But for simple shell script substitution we usually change
LOCAL_CPP_DEFINES = -DCDE_CONFIGURATION_TOP=$(CDE_CONFIGURATION_TOP) \
-DCDE_INSTALLATION_TOP=$(CDE_INSTALLATION_TOP) \
-DCDE_LOGFILES_TOP=$(CDE_LOGFILES_TOP)
to:
LOCAL_CPP_DEFINES = -DCDE_CONFIGURATION_TOP=$(CDE_CONFIGURATION_TOP) \
-DCDE_INSTALLATION_TOP=$(CDE_INSTALLATION_TOP) \
-DCDE_LOGFILES_TOP=$(CDE_LOGFILES_TOP) \
-DKORNSHELL=$(KORNSHELL) \
-DXPROJECTROOT=X11ProjectRoot
since we don't want quotes for shell scripts.
Fix warnings related to secruity concerns on varargs functions. By specifying
"%s" on single string calls to sprintf() (and related) it's not possible to
have a % in the input string causing random data to be read off the stack.
tmpnam() usage replaced with mkstemp(). Find a suitable tmp directory
checking the TMPDIR environment variable first, then the P_tmpdir
macro and finally /tmp directly.
On 64-bit Linux platforms, check to see if libc.so exists in /usr/lib64.
If found, use it over /usr/lib/libc.so.
The libc.so file is not always in /usr/lib. On multilib systems, the
file we care about could be in /usr/lib64. Likewise, common Linux
conventions call for 64-bit libraries to go in lib64 directories, so
check there first when on a Linux 64-bit system.
Use the same set of langs as on Linux and FreeBSD (no Japanese), don't
redefine a needed macro as no-op, and unset LC_CTYPE in the environment
when building cat files.
We need to use mkcatdefs to build those.
mkcatdefs needs to be built from the open motif
source tree (localized/util/mkcatdefs.c)
and installed as:
cde/imports/motif/localized/util/mkcatdefs
GetBaseName causes segfaults, because when the pathname
ends in a "/" it returns NULL. This happens when trying
to give a valid filename to dtcreate for an icon.
ProcessExecString thought it was returning an array of size 3; however
in C arrays are second-class and there is no direct way to return an
array like this; GCC warning triggered because it was actually
just returning a pointer to local storage. Fixed using malloc.
Also fix some obviously wrong usages of sizeof, although they were
relatively harmless. A little other warning quieting using 0 instead of
NULL.
This code always buffer overflowed, because exactly 2 bytes
less than were used were allocated. This led to dtcreate
crashing when hitting "Find Set..."
The following font families
(or their aliases) will be used:
-adobe-courier-bold-o-normal--*-
-adobe-courier-bold-r-normal--*-
-adobe-courier-medium-o-normal--*-
-adobe-courier-medium-r-normal--*-
-adobe-helvetica-bold-o-normal--*-
-adobe-helvetica-bold-r-normal--*-
-adobe-helvetica-medium-o-normal--*-
-adobe-helvetica-medium-r-normal--*-
-adobe-symbol-medium-r-normal--*-
-adobe-times-bold-i-normal--*-
-adobe-times-bold-r-normal--*-
-adobe-times-medium-i-normal--*-
-adobe-times-medium-r-normal--*-
-b&h-lucidasans-medium-r-normal-sans-*-
-b&h-lucidatypewriter-bold-r-normal-sans-*-
-b&h-lucidatypewriter-medium-r-normal-sans-*-
The files will be installed in
/usr/dt/config/xfonts/C
This directory should be added to the
X server font path:
xset fp+ /usr/dt/config/xfonts/C
and/or via
FontPath "/usr/dt/config/xfonts/C"
in the "Files" section of the xorg.conf file.
We have
pid_t
wait3(int *status, int options, struct rusage *rusage);
on FreeBSD and we don't need (union wait) handling.
Another good candidate for one #ifdef from imake templates.
We have already OPT_BSD_WAIT in ToolTalk's tt_options.h
Add preprocessor directives not to try
to redefine sys_errlist[] or sys_nerr
There are already definitions:
extern __const char *__const sys_errlist[];
extern __const int sys_nerr;
in <stdio.h>
Actually we should have something like
NeedSysErrlist in imake definitions
to get rid of those #ifdefs.
This code tried to automatically generate the X DISPLAY
from the combination of the hostname and display number;
however 127.0.0.1:0 is normally rejected by X11, so this
technique is no good. Fixes dticon hang on startup, caused by
XOpenDisplay failure leading to this message from tttrace:
tt_default_session_set(0x0x875190=="X 127.0.0.1 0") = 1032 (TT_ERR_ACCESS)
BSD make interrupts shell pipeline after
if it cannot run the command:
rm -f Mrm.msg
ln -s ../../../../imports/motif/localized/de_DE.ISO8859-1/msg/Mrm.msg Mrm.msg
Running mkcatdefs for Mrm.cat with LANG set to de_DE.ISO8859-1
( rm -f Mrm.cat Mrm.tmp.msg; LANG=de_DE.ISO8859-1; export LANG; ../../../../imports/motif/localized/util/mkcatdefs Mrm Mrm.msg -h > Mrm.tmp.msg; gencat Mrm.cat Mrm.tmp.msg; rm -f Mrm.tmp.msg )
../../../../imports/motif/localized/util/mkcatdefs: not found
*** Error code 127
After this, an empty Mrm.tmp.msg is left.
- Const strings referenced by non-const variables.
- Incorrect format specifers for printing addresses
- Unused variables
- Signed comparison to unsigned
Also fix an incorrect enumeration value in a switch statement.
Introduce KORNSHELL make variable to point
to the implementation of the Korn Shell.
Use $(SHELL) or $(KORNSHELL) explicitly for
make programs that do not automatically call
shell scripts from the current directory.
dtksh can be now compiled on FreeBSD. Work in progress.
Needs a real Korn shell to bootstrap as $(KSHELL).
KSHELL is set by default to /usr/local/bin/ksh93
(generic POSIX shell may not work)
Tested on:
FreeBSD 9.0-BETA1 #0 r224912M amd64
Known issues:
xvmstat:
* sleep does not work well (SIGSTOP is delivered)
xpong:
* xpong: line 220: ball1x = max_x * 2.2 / 3 : arithmetic syntax error
* dtksh is rebuilt uncondtionally every time make is invoked
Fixes many, though not all 64bit-warnings. In lots of places, pointers are
cast to ints to be then used as array subscripts. The only way to deal with
this is to change them to long. Additionally, use calloc() to allocate the
int_array in istr.c and drop the (wrong) macro patch to istr.h. Should make
dtbuilder work on 32bit again.
Fixes:
agent.c: In function '_DtCm_init_agent':
agent.c:160: warning: passing argument 5 of 'registerrpc' from incompatible pointer type
agent.c:160: warning: passing argument 6 of 'registerrpc' from incompatible pointer type
agent.c:167: warning: passing argument 5 of 'registerrpc' from incompatible pointer type
agent.c:167: warning: passing argument 6 of 'registerrpc' from incompatible pointer type
This reverts commit 0d2f7866ac.
This causes great mayhem in building/generating dtbuilder .msg files
(corrupting them, and inserting '(nil)' all over the place).
These would cause dtbuilder, and any other program built by dtcodegen
to have screwed up colors, missing callbacks and other mayhem.
This was confirmed by others on the list - reverting this made those
issues go away.
It may be that the int -> long is correct, but the NULL check
certainly does not seem to do what was intended. I'll leave it up to
Pascal to investigate :)
cpp was removed from /usr/libexec with FreeBSD 5.0-RELEASE.
Use __FreeBSD_version to tell imake if it's still there.
While here, X.Y.Z versioning ended with FreeBSD 3.0, so
we must be dealing with 2.Y.Z when checking for -lgnumalloc.
A patch from Pascal Stumpf using external jpeg broke linux builds,
since libjpeg needs to be linked in when using a remote jpeg lib.
So, in lnxLib.tmpl, define SharedDtHelpReqs so -ljpeg is used. Also,
in DtHelp/Imakefile, use proper Arch defines so external jpeg libs are
only used on linux, fbsd, and obsd systems.
This is a non-POSIX/ISO-C header. It is ok to include this on Linux, but it
is obsolete on BSD; FreeBSD even throws an error if you include it with
__STDC__ defined. Every system should nowadays have malloc() defined in
stdlib.h.
Diff is largely mechanical, replacing malloc.h with stdlib.h where it is not
yet included anyway.
Nowadays, OpenMotif is itself linked to libjpeg, so pulling in another
version of it causes symbol size mismatches, not to mention the maintenance
burden and security implications arising from keeping our own copy of libjpeg.
We still need some of the header files provided here because they are internal
to libjpeg and not installed on most distributions.
In part of the tooltalk rpc code (mp_message.c), it was assumed that on
the majority of platforms, sizeof(uid_t)=sizeof(gid_t)=sizeof(long). On
Linux-x64, uid_t is an unsigned int, which makes the code fail: all
tooltalk messages fail to send with an RPC_CANTENCODEARGS at the
rpc-level, and TT_INTERNAL_ERR for the actual program. We instead
change the code to explicitly examine sizeof(uid_t) to see whether it is
int or long sized. This allows tooltalk-dependent functinoality
like logout and multiple calls to dtfile to work.
Enums may be represented with a smaller type than int; however, they are
automatically promoted to int when passed in va_arg lists, just as
short, char, etc. are. GCC thus "knows" that you never want to call
va_arg with an enum type, and instead inserts an abort.
strstream.h is now called "strstream" and is obsolete, but use it anyway until
all code is converted over. This also needs std:: added, at least for GCC
4.2.1. Lastly, when hardcoding the path to perl, /usr/bin/perl should be used
rather than anything else.
Some of these older files just clutter up things and do not contribute
much information that is not historical in nature, so preserve them
in an out of the way location.
Mostly this is adding appropriate #includes and declarations,
but for WmImage.c we also change from using the proper name
for XmeGetMask, rather than the identical but renamed
version _DtGetMask which is not exported in any header.
Patch from Pascal Stumpf <Pascal.Stumpf@cubes.de>
The official POSIX name for this signal is SIGCHLD. Linux probably
has SIGCLD only for SysV compatibility, but BSD does not.
Patch from Pascal Stumpf <Pascal.Stumpf@cubes.de>:
So here are all the patches that deal with the fact that modern
compilers assume different scoping rules for variables declared in for
loops. On Linux, -fpermissive has been added as a compiler flag to
compensate for this old C code, but I think it is the wrong approach.
Sorry, couldn't help sneaking in a || defined(CSRG_BASED) and some casts
needed for other reasons ...
Patch from Pascal Stumpf <Pascal.Stumpf@cubes.de>
Most is just copied from the existing imake installation in
/usr/X11R6. Additionally:
* Allow overriding CDESharedRev, X11ProjectRoot and ProjectRoot
* Add a new define to be able to override MINCLUDESRC too
(this allows me to build CDE without symlinking any include directories)
don't use the idiom
char foo[BUFSIZ];
snprintf(foo, BUFSIZ, ....);
but
char foo[BUFSIZ];
snprintf(foo, sizeo foo, ....);
because this will automatically catch situations where the size of foo
is later changed, e.g. like foo[BUFSIZ + 8];
Fix another use of sprintf.
Patch from Robert Tomsick <robert+cde@tomsick.net>:
I believe this fixes vulnerability #3 from CERT CA-1999-11.[1] The other
uses of sprintf in DtAction seem to be safe.
[1] https://www.cert.org/advisories/CA-1999-11.html
Patch from Douglas Mencken <dougmencken@gmail.com>:
"%wc" is Microsoft extension, not supported in every Std C Library. So
if we don't want to print "%wc%wc%wc%wc%..." instead of real chars, we
shall not use it.
Before:
%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wcession[28326]:
_Tt_s_session::s_init(): 1051 (TT_ERR_INTERNAL)!
%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wcession[28326]: waitpid():
No child processes
%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wc%wcession[28326]: child
ttsession exited with status 1
After:
/usr/dt/bin/ttsession[12397]: _Tt_s_session::s_init(): 1051 (TT_ERR_INTERNAL)!
/usr/dt/bin/ttsession[12397]: waitpid(): No child processes
/usr/dt/bin/ttsession[12397]: child ttsession exited with status 1
Patch from Frederic Koehler <f.koehler427@gmail.com>:
These implicit definitions cause segfaults on x64 because
the implicit return type is a 32-bit signed int, rather than a pointer
type.
Patch from Frederic Koehler <f.koehler427@gmail.com>:
Define a final fallback for loading default window manager font;
before exiting, forcefully try to load "fixed" font. This is sufficient
to allow systems where fontList is set to an empty list to startup dtwm,
for now.
dtlogin's genauth routines were trying to open and read /dev/mem on
linux and (presumably) bsd systems in order to obtain random data used
in creating an auth key.
This is bad for a variety of reasons. Newer linux kernels (at least
on 3.2) issue the following warning to the kernel logs:
"Program dtlogin tried to access /dev/mem between 100000->102000."
Now on linux we will use /dev/urandom, and on CSRG_BASED (bsd) systems
we will use /dev/random to obtain some entropy.
With Aaron's fixes to dtdbcache fixing a potential coredump, the
comment block in the write_db() function regarding tmpnam() no longer
applies, and the tmpnam_buf variable is no longer used.
So, remove them :)
Patch from Ulrich Wilkens <mail@uwilkens.de>
I have a little patch for a problem that I found when I tried to
compile dthelp on 64bit FreeBSD. It could also be a problem on other
64bit systems. The problem is that the program context compiles but
fails running with segmentation fault.
context uses the function m_malloc() which is missing a correct
prototype sometimes. Then it's treated to return int instead of void *
. On 64bit systems this cuts off the higher 32 bits because void * is
64bit whereas int is only 32bit.
Patch from Douglas Mencken <dougmencken@gmail.com>:
The issue is that MESSAGE tries to invoke catgets with NULL first
parameter, which is dereferenced inside catgets (Std C Library
function) without checking, from catgets.c source:
if (catalog->name_ptr[idx + 0] == (u_int32_t) set
&& catalog->name_ptr[idx + 1] == (u_int32_t) message)
On the other hand, there's a special value: -1 (cast to nl_catd),
which must be used instead of 0 (NULL) in the case when we are unable
to provide real catalog_desc, from catgets.c source:
/* Be generous if catalog which failed to be open is used. */
if (catalog_desc == (nl_catd) -1 || ++set <= 0 || message < 0)
return (char *) string;
In dtlogin, you can select the language to switch to by selecting it
via Options->Language. Unfortunately this was also including '.' and
'..', since this list is built dynamically by scanning a directory.
Now we screen out '.' and '..'.
The TOG copyrights were being removed after a rebuild, leaving behind
the scary "RESTRICTED" copyright text that was originally there.
The issue was that the TOG copyrights were not properly embedded
within a 'DTB_USER_CODE_START' code block.
dtcodegen does not preserve any code outside DT_USER_CODE START and
END blocks.
Additionally, these objects are built with -merge by dtcodgen, so the
existing 'RESTRICTED' header within the codeblocks was being retained.
It is amazing what mayhem can be caused by a bad define :)
The reason most of the CDE programs were not actually using their
localized catalogs was because their use was being disabled by this
line.
Certain programs like dtwm and dthelp/libDtHelp did their own
localization handling and did not use the results of this define.
This is why they worked properly, and most everything else did not :)
/usr/share/terminfo as well as /usr/share/lib/terminfo
This allows fully functioning dtterm on Ubuntu 12.04 on which the terminfo file
failed to install.
The dtwm DefaultWindowMenu did not list any of the workspace enabled
Occupy commands, which makes it a bit difficult to move windows into a
workspace, or to occupy multiple workspaces.
This was rectified by copying the relevant lines from the
SampleWindowMenu sections containing them that already existed in the
localized files. Tested on C and DE (German).
instant should fail if the locale xlation db cannot be opened.
The reason this was failing was because the locale translation DB for
linux did not exist (Linux.lcx). I didn't add it until weeks/months
later :)
There needs to be one for each platform called "$(uname -s).lcx".
FreeBSD will need one too.
Note, there are still some issues with certain programs appearing not
to use their localized message catalogs properly, while others work
fine.
Also, in order to build these now, you need to make sure you have
installed the DE, ES, FR, and IT locales before building or you will
get failures. On [k]ubuntu, at least, you can install these with the
following commands:
sudo locale-gen de_DE
sudo locale-gen es_ES
sudo locale-gen fr_FR
sudo locale-gen it_IT
In order for this to work, dtsession must be setuid root. If
dtsession is not setuid root, then locking will be disabled, and a
message will be written to ~/.dt/errorlog with the message: "Unable to
lock display due to security restrictions".
Note, this requires that the xfonts-100dpi and
xfonts-100dpi-transcode[d] packages be installed. Kubuntu 11.10 calls
the transcoded package 'xfonts-100dpi-transcoded' while 12.04 calls it
'xfonts-100dpi-transcode'
You can also use the 75dpi variants if you wish, though they will look
crappy on larger monitors (>1024x768).
What we really need are more fonts installed, like all of the xf 75dpi
and 100dpi fonts. 100dpi looks much better than 75/72 dpi, especially
on any display larger than 1024x768. Of course, in the far future, we
should use the anti-aliased TT fonts everyone else uses these days
anyway.
Note, these still aren't quie working yet. Notably, font.dir needs to
be generated properly and re-committed so they will be used.
mkfontdir is used to do this, but currently it fails on these
font.alias files (does not recognize them).
Use the help system instead. Someday, if dtinfo can be made to work,
this can be undone easily. Note, you may have to clear out your ~/.dt
dir to see the change.
Under ubuntu at least, /bin/sh is really /bin/dash. For some reason,
dash cannot detect an executable (-x) file if it resides on nfs. bash
and ksh do not have this problem. Since ksh is already required to
build and install major portions of CDE, might as well use it here too.
Added proper SharedDtSvcReqs in lnxLib.tmpl and CplusplusLibC in
linux.cf. This allows the libstdc++ dependancy to be properly
declared for libDtSvc so that it is not neccessary to hardcode 'CCLINK
= g++' in the Imakefiles of programs linking angainst libDtSvc.
- by default, do not build any other locale than C for now
- do not try to build the guides. These require functioning
dtinfo/docbook
- add a Linux.lcx locale translation db. Not used yet.
- fix some overflows in dtdocbook/instant
- also some rework of linux.cf
- only allow ELF systems
- use -pipe
- add some more defines
- get rid of some of the ancient (libc5/linux 1) support. Really,
don't expect CDE to build right on such old systems.
- databases need linux specific entries for CDE-SHLIBS
- database Imakefile should undef the platform define (linux)
- installCDE fixes
- use $LOGFILE. There are still cases where tmp files are created
in /tmp.
- detect and set proper PLATFORM
- Check to see if a db exists before trying to use it
- don't print usage if you aren't root.
Remove from individual Imakefiles.
Also, remove '#if 0' block in linux.cf, and remove empty
LinuxMachineDefines. This should be working correctly. If not, let me
know.
Do not require the user to be root
Use -e in echo statments so that embedded \t's etc will display.
Do not save install log in /tmp with a known name, especially since
it is supposed to be run as root. This is a Bad Thing To Do (tm).
Instead just save the log in the current dir.
